PHD THESIS DANISH MEDICAL JOURNAL
This review has been accepted as a thesis together with 1 previously published and 2 submitted papers (have since been published) by Aarhus University on 9 October 2014 and defended on 16 January 2015.
Tutor(s): Per Fink, Lisbeth Frostholm & Winfried Rief
Official opponents: Arthur Barsky, Alexandra Martin & Henrik Kolstad
Correspondence: Department, The Research Clinic for Functional Disorders and Psychosomatics, Aarhus University Hospital, Noerrebrogade 44, 8000 Aarhus C, Denmark
E-mail: trieil@rm.dk
Dan Med J 2016;63(10):B5294
THE 3 ORIGINAL PAPERS ARE
Paper I. Eilenberg T, Kronstrand L, Fink P, Frostholm L. Ac- ceptance and commitment group therapy for health anxiety - Results from a pilot study. J Anxiety Disord 2013 Jun 19;27(5):461- 8.
Paper II. Eilenberg T., Fink P., Jensen JS, Rief W., Frostholm L.
Acceptance and Commitment Group Therapy (ACT-G) for health anxiety: A randomised, controlled trial. Psychological Med 2015;46(01):103-15.
Paper III. Eilenberg T, Frostholm L, Schröder A, Jensen JS, Fink P.
Long-term consequences of severe health anxiety on sick leave in treated and untreated patients: Analysis alongside a randomised controlled trial. J Anxiety Disord 2015;32:95-102.
SETTING AND OUTLINE OF THE THESIS Outline
The present thesis examines whether group therapy combining well-known cognitive-behavioral strategies with an acceptance- based approach (Acceptance and Commitment Therapy in groups, ACT-G) improves symptoms of illness worry (paper I) compared to a waitlist condition in patients with severe health anxiety (paper II). Furthermore, it is investigated whether persons with untreat- ed severe health anxiety a) have higher levels of sick leave com- pared with the general population, and b) experience a decrease in sick leave in the year following ACT-G treatment (paper III).
Throughout the thesis, the term health anxiety is used synony- mously with designations such as hypochondriasis,
abridged hypochondriasis, hypochondrical disorder and illness anxiety disorder, unless differences between the diagnostic labels are discussed.
The thesis begins with a general introduction to health anxiety with special emphasis on classification of health anxiety and Perspectives on essential features and psychological treatment of health anxiety. The aims of this thesis are presented after that introduction.
The thesis deals with treatment of severe health anxiety and describes a pilot study testing the feasibility and acceptance of ACT-G for severe health anxiety (paper I) before initiating a ran- domised controlled trial (RCT). It continues to describe how ACT- G was implemented in an RCT and evaluates the treatment effect on illness worry and secondary outcomes of emotional distress, physical symptoms and health-related quality of life as well as the level of acceptance of the research diagnosis health anxiety 1 (paper II). The treatment manual can be requested from the author.
The thesis also concerns the influence of severe health anxiety on sick leave compared with a matched general population sample and explores the one-year treatment effect of ACT-G on sick leave among patients with severe health anxiety (paper III).
As the primary objective of this PhD project was conducting and evaluating the RCT trial, the general discussion on methods and results in this thesis will predominantly be focused on aspects of the RCT 4.
The thesis continues with a general discussion on methods and a summary of results in relation to the aims of the thesis followed by an overall discussion of results, perspectives for further re- search, English and Danish summaries and a reference list.
Setting
Paper I provides original data from a pilot study (n=34) carried out at the Research Clinic for Functional Disorders and Psychosomat- ics, Aarhus University Hospital, Denmark between 2009 and 2010.
Paper II provides original data from 126 patients enrolled in an RCT carried out at the Research Clinic for Functional Disorders and Psychosomatics, Aarhus University Hospital, Denmark be- tween 2010 and 2012 (the trial is registered at clinicaltrials.gov, no. NCT01158430). Designing, conducting and evaluating the RCT trial were the primary objectives of this PhD project.
Paper III provides original data from a national database on sick- ness-related benefits from the 126 patients enrolled in the RCT
Acceptance and Commitment Group Therapy (ACT-G) for health anxiety
A randomised, controlled trial
Trine Eilenberg
and a matched general population sample of 12,600 individuals and assesses the effect of severe health anxiety on sick leave.
INTRODUCTION Aims
A brief overview is given of how health anxiety has been concep- tualised and classified over the years. The aim of the overview is to identify some of the key issues with respect to classification and labelling of the disorder that have caused major obstacles for research and may have hindered effective patient care 5. The introduction will initially present two vignettes to illustrate the multifaceted phenomenon of health anxiety. Furthermore, a brief overview of essential features of severe health anxiety will be outlined in order to give a better understanding of how health anxiety is currently treated. Finally, the new treatment approach Acceptance and Commitment Therapy (ACT) for health anxiety is described.
What is health anxiety?
Health anxiety may be seen as a multifaceted phenomenon rang- ing from mild and transient to severe and chronic conditions 6. Most people have experienced transient health anxiety at times when experiencing unpleasant and persistent symptoms. For some people, worry and rumination about illness becomes a maladaptive behaviour as the rumination and perceived risk are out of proportion with the objective degree of medical risk and hence cause great personal suffering and widespread impair- ment. The vignettes serve to illustrate two examples of patients both suffering from impairing health anxiety, but who show very different behavioural patterns.
The behavioural patterns seen in the cases may be categorised as respectively care-seeking and care-avoidant behaviour 7. Yet, these patterns are not thought of as definitive characteristics of severe health anxiety, other characteristics could just as well have been highlighted. Primarily, the two cases are chosen to illustrate that patients with severe health anxiety may present quite differ- ently, but still share the excessive ruminations with intrusive worries about harbouring serious illness and a persistent preoc- cupation with health leading to significant impairment. In the present thesis, the focus will be on patients suffering from severe and impairing health anxiety.
Diagnostic classification of health anxiety
The term hypochondriasis is for most people considered a stigma- tising label that has a pejorative connotation as it implies that the symptoms are not real and that it is all “in your mind”, bordering on faking 8. Therefore, health anxiety has been suggested as replacement, and in the present thesis, this designation and diagnostic criteria will be used.
The view of health anxiety has changed significantly throughout times, yet the term hypochondria has been in use since ancient times. In 460 BC Hippocrates referred to it as the anatomical region that is located under the curvature and it was assumed that hysteria was due to a migrant uterus 9. From around the 19th century and until a renewed interest arose in the 1960s, this condition received scarce attention in the literature as hypochon- dria was considered a manifestation of neurasthenia 10. Hence, hypochondriasis was first included in the official classification system of DSM-II in 1968 11 as a neurosis under the designation hypochondrical neurosis and with a focus on bodily preoccupation and fear of having a disease. In 1980, with the introduction of the DSM-III 12, it was added that the preoccupation and fear was characterised by unrealistic interpretation of physical signs and sensations and that these symptoms could not be explained by another disorder, nor could the patient’s fear or belief be dis- missed through repeated medical reassurance. In the DSM-III edition, hypochondriasis was now moved to the new diagnostic group of somatoform disorders including a group of mental disor-
Vignette 2.
Ann is a 55-year old married woman working in a sales office.
Ann starts training as a nurse at the age of 20. During her training as a nurse she becomes increasingly occupied by sensations in her body. More and more often when she in her training is confronted with symptoms of illness, she experienc- es the same symptoms herself. Ann is aware of this pattern and finds it very humiliating; still she is unable to control her rumination about illness. After one year of training as a nurse she decides to drop out. At first Ann finds a relief in illness worries, but in order to control her worries she has to perform a ritual of self-examination several times a day. Also, Ann finds that if she keeps herself very occupied all day she can keep the worries at a distance. To be able to sleep at night Ann has to take sleeping pills. Though Ann is severely impaired by her illness worries, and even unable to be in a relationship because of this, she never talks about her illness worries or have any contact with the health care system. She does not respond to the preventive check ups (e.g. mammography, smear test) offered in the health care system as she is too afraid of the results.
Vignette 1.
Peter is a 45-year-old academic who suffers from a strong, bothersome inclination to always be afraid of being seri- ously ill. It is especially cancer that he fears. Peter remem- bers from his childhood that both his mother and grand- mother often expressed concerns regarding illnesses.
Peter’s worries regarding illness changed dramatically as he himself became a parent at the age of 30. From this time whenever Peter is aware of the slightest change in bodily sensations, that be a headache or a mark on the skin, he is unable to concentrate on any task till his general practitioner (GP) reassures him that his symptoms are harmless and normal. Also, Peter is a frequent user of medical homepages trying himself to classify his symp- toms. Peter’s social relationships are burdened by his worries as he brings up topics regarding illness in most conversations. Over the years Peter experience that reas- surance from his GP had less positive impact on his worries – he can no longer believe it when his GP tells him that he needs no more tests and that there is no need to worry.
Even though, Peter no longer experiences a relief in worry when consulting his GP, he still frequently contacts him.
Lately, Peter has made an appointment with a private hospital in order to get a brain scan, as he fears he has a brain tumour and his GP will not refer him to any more tests.
ders that predominantly manifested themselves as physical symp- toms. In the revised DSM-III 13, a duration criteria of minimum 6 months was added. The definition of hypochondriasis was further refined in the DSM-IV edition 14 in 1994 in which specific exclu- sionary criteria were added in order to better discriminate hypo- chondriasis from other disorders (e.g anxiety disorders, depres- sive disorders or other somatoform disorders) as well as the presence of significant distress or functional impairment.
In cases where the full DSM-IV criteria for hypochondriasis are not met, due to one or more of the diagnostic features not being present, the condition has been called abridged hypochondriasis
15.
In recent years, the DSM-IV hypochondriasis diagnosis has been criticised for being too restrictive 15, 16 resulting in too low preva- lence rates 17 and neither satisfying clinical nor nosologic validity requirements 1. The empirical foundation for the DSM-IV criteria has been found poor as it has arisen from mainly clinical observa- tions of patients in severely skewed psychiatric settings, even though the disorder is predominantly seen in medical settings 1. Especially criterion B regarding “appropriate medical evaluation and reassurance” has been criticised 1, 15 as some patients delib- erately avoid their general practitioner (GP) as a maladaptive avoidance (such as the patient in Vignette no 2) or due to dissatis- faction with previous health care experiences. Furthermore, criterion E regarding duration of symptoms for at least 6 months has been argued to be arbitrary and restrict the diagnosis to a chronic sample 1, 18.
In 2004, Fink et al. 1 introduced new and empirically-based positive diagnostic criteria for health anxiety, in which severe health anxiety is characterised by exaggerated rumination with intrusive worries about harbouring serious illness and a persistent preoccupation with one’s health leading to significant impairment and a decrease in quality of life (see Table 1.1 for the research criteria for severe health anxiety). According to the new diagnos- tic criteria, health anxiety should no longer be a diagnosis of exclusion and may be helpful in providing patients with a positive explanation of their symptoms. In the trials of this PhD project, these diagnostic criteria have been used for inclusion.
Table 1. Diagnostic criteria for severe health anxiety 1
This year, new diagnostic criteria for health anxiety have been introduced in DSM-V 7, in which patients with former DSM-IV hypochondriasis are now subsumed under the classification of
Illness anxiety Disorder (excludes patients with moderate and severe somatic symptoms) or Somatic Symptom Disorder (SSD).
It has recently been suggested that using the broadly defined DSM-IV hypochondriasis diagnostic criteria may result in a very heterogeneous sample 1, 19 resembling the SSD criteria of the DSM-V more than that of the Illness anxiety disorder 20. The diag- nostic health anxiety criteria used in the present thesis are found to be rather similar to the DSM-V Illness anxiety disorder, the difference between the two primarily being that the DSM-V diag- nosis does not include the rumination symptom which is the key criteria of the health anxiety diagnosis, and furthermore the DSM- V diagnosis contrary to Fink et al.’s 1 excludes patients with mod- erate and severe somatic symptoms.
Overall, the lack of empirically valid diagnostic criteria have re- sulted in lacking expert consensus on how to classify or label patients with severe and persistent illness worry, which have hampered research and hence may have blurred the effect in trials of treatment for health anxiety. It is difficult to compare effects of treatment between studies when very different diag- nostic labels and criteria have been used for inclusion. Further- more, comparison between studies are also challenged by some studies identifying cases of health anxiety based on subjective scores on self-reported questionnaires on illness worry (with arbitrary cut-off criteria) compared to using diagnostic criteria for inclusion. This may potentially further blur the characteristics of the sample.
Boundaries of health anxiety and comorbidity
Like many clinical syndromes, health anxiety has characteristics that cross boundaries and suggest a shared aetiology with espe- cially other somatoform disorders and anxiety disorders 16. Health anxiety may be said to present with mixed symptoms of anxiety, such as rumination and catastrophic thinking, and somatic symp- toms. Somatisation disorder and health anxiety share the pres- ence of physical symptoms for which no organ-pathology can be found. Though, there is a difference in the way patients relate to symptoms and what is the main source of their distress. The majority of patients with somatoform syndromes experience that it is the somatic symptoms per se, such as unexplained pain, spasms, and fatigue, that cause distress, whereas for health anxi- ety, illness worry itself (rumination about the meaning, implica- tion and consequences of the symptoms) is the primary problem suggested to arise from misperceptions of innocuous somatic symptoms 21.
Although health anxiety remains classified as a somatoform dis- order in the new DSM-V 7, considerations regarding alternative classification as an anxiety disorder were made 22 emphasising the conceptual relationship between health anxiety and anxiety disorders. Also, some evidence for phenomenological similarities between health anxiety and anxiety disorders have been shown
23-25. A study comparing patients with hypochondriasis, somato- form disorders and anxiety disorder respectively on sociodemo- graphic variables, psychopatology and treatment effects found that patients with health anxiety have an interim position be- tween the two disorders, but is slightly closer related to anxiety disorders 26. Due to the chronic course seen in health anxiety and the pervasive impact on behaviour and cognition, some 27 have also suggested that the disorder may be better classified along Axis II as a personality disorder. The personality characteristic primarily found to be related to health anxiety is neuroticism 28,
29, and recently a long-term follow-up study has looked at per- sonality, measured by an inventory of dimensions of tempera- ment and character as a predictor for remission after treatment.
Key criteria: Rumination with intrusive thoughts and ideas, and fears of harbouring an illness
At least 1 of 5 sub-criteria;
1) a) Worries, preoccupation or fear of harboring a severe physical disease
b) Attention to an awareness of bodily functions 2) Suggestibility or autosuggestibility
3) Excessive fascination with medical information 4) Fear of being infected or contaminated 5) Fear of taking prescribed medication
Mild or severe according to influence on functioning and well-being
Duration more than 2 weeks.
The study found preliminary evidence that patients with health anxiety, who have a less harm-avoidant temperament, for in- stance being less fearful and vulnerable, and a more cooperative character, such as more tolerant and forgiving, were associated with shorter remission time and full remission at follow-up 30. The literature has greatly debated whether health anxiety should be considered primary or secondary to other diagnoses such as depression, and it is now at large recognised that health anxiety can occur without past or current conditions 31. Health anxiety is seen to co-occur with numerous psychiatric disorders, and more than half of patients with health anxiety may have a comorbid psychiatric disorder, the most prevalent being major depression or anxiety disorders 1, 32. Likewise, studies have found that be- tween 1/5 and 1/3 of patients with health anxiety also fulfil diag- nostic criteria for another somatoform disorder, the most preva- lent being undifferentiated somatoform disorder and with the highest overlap seen among patients with full DSM-III or IV hypo- chondriasis 1, 32. Although studies show that around half of pa- tients with health anxiety suffer from comorbid anxiety and de- pression, psychiatric comorbidity does not seem to predict the course of health anxiety or prevent patients with health anxiety from improving 5, 33, 34.
Prevalence of health anxiety
Health anxiety is common, yet prevalence estimates vary consid- erably across studies. A recent review 17 of 55 papers based on 47 independent samples concluded that prevalence rates across the included studies were difficult to compare due to the heteroge- neous nature of the studies such as use of different definitions and assessments and different populations, and therefore it is questionable whether reported prevalence rates reflect the same phenomena. Though, the review found a trend towards a signifi- cantly higher population prevalence rate in abridged forms of hypochondriasis (e.g. prevalence range for health anxiety was 2.1- 13.1%) compared to full DSM-IV hypochondriasis (weighted prev- alence of 0.40%, range 0.0-4.5%). Also, prevalence rates were found to be higher in general medical samples (prevalence of full criteria; range 0.3-8.5% and for health anxiety: range 4.5-30.6%) compared to clinical samples (psychiatric, psychotherapy and psychosomatic settings: range 0.4-7.4%). Due to the heterogene- ous assessment methods and diagnostic definitions, calculated prevalence rates must be interpreted with caution.
The aetiology of health anxiety
Research in the aetiology of health anxiety is increasing, and a number of aetiological factors have been the focus of research:
Family background and childhood experiences, stressful life events, sociocultural factors and genetic factors. So far most studies have failed to find direct evidence for genetic factors in health anxiety, and behavioural genetics suggest that health anxiety is moderately heritable, but is more strongly influenced by environmental factors 35. A study 36 investigated the role of genetic and environmental factors using a classic twin study method and found genetic factors to account for between 10-37%
of the variance of excessive health anxiety. The authors conclud- ed that health anxiety is largely a learned phenomenon and hence support the use of environmental interventions such as psycho- therapy as treatment for health anxiety. Retrospective studies have shown that environmental factors such as traumatic child- hood experiences, e.g. childhood abuse 37, 38 and parental model- ling of illness worry as well as sensitivity to somatic sensations 39 are associated with health anxiety. Still, the exact contributions of environmental and genetic factors to the development of health
anxiety are still largely unknown. It might be that genetic factors cause a proneness or sensitivity towards experiencing negative emotions or bodily sensations. Also, positive associations seen between environmental factors and health anxiety are most likely complex. For example, a recent study 38 on adverse childhood experiences and health anxiety in adulthood showed a significant positive association between the two and found that childhood experiences were predictive of health anxiety in adulthood, yet the unique contribution of these experiences lost significance when other variables of interest such as negative affect and trait anxiety were included in the analyses.
A better understanding of the aetiological factors in health anxie- ty might help guide preventive approaches (e.g. during childhood) or develop better targeted interventions. Also, the above men- tioned twin study 36 stressed that as genetic factors play some role in excessive health anxiety, it may be that pharmacotherapy can be improved in the future by tailoring medications to the person’s genotype.
Onset and course
Health anxiety may arise at any age, although an early onset has been suggested40. In respect to the natural course of health anxi- ety, a systematic review of the epidemiology of hypochondriasis found that there are inadequate longitudinal studies allowing for exact determination of these factors 19. Still, a number of studies have found an increased risk of chronic course in severe cases of health anxiety left untreated 5, 41-43. Most studies have found no gender or age differences in the prevalence of health anxiety 1, 15,
44-46, whereas within other somatisation disorders women seem to be at higher risk 19. In general, RCTs have found that patients with severe health anxiety are well-educated with approximately 2/3 of patient samples having attended further education 47-51. Furthermore, a chronic course of health anxiety has been associ- ated with more severe symptoms, more impaired physical func- tioning, childhood punishment and a longer duration of health anxiety at baseline 52, 53.
Personal and socio-economic costs of health anxiety
Severe health anxiety can cause frustration in both GPs and pa- tients. GPs may find it frustrating that they cannot offer the same treatment quality as they do to other patients 54. Also, some GPs may be less comfortable addressing health anxiety directly com- pared to offering information on physical health and therefore rarely diagnose health anxiety, and some may also be inclined to view health anxiety as non-responsive to therapy or even as not a genuine disorder. Besides struggling with substantial personal suffering, patients with severe health anxiety may experience a lack of sufficient help and that their problems are not taken seri- ously. Also, patients with severe health anxiety may be reluctant to be referred from a medical setting to a psychiatric setting 55. A two-year follow-up study in primary care showed that health anxiety is a persistent condition and spontaneous remission is rare
5. Furthermore, the study found that patients with severe health anxiety used about 41-78% more health care per year in total, both during the 3 years preceding inclusion and during follow-up, com- pared with patients with well-defined medical conditions 5. Pa- tients with severe health anxiety have a high use of health care services as they go through a wide range of examinations, assess- ments and treatment attempts without medical indication 56-59. Earlier findings demonstrate that training GPs in management and treatment of patients with functional disorders can improve the treatment quality 60, yet this does not seem to be the case in the
management of patients with health anxiety. Results indicate that severe cases demand a more intensive and specialized treatment 5,
61.
Due to the numerous investigations made to rule out any medical condition as well as reassurance-consultations in primary care, health anxiety is expensive in terms of direct health care resource usage 5, 62-64. Yet, the high health care costs do not include time lost from work and reduced productivity. Studies on indirect costs such as morbidity-related sick leave or productivity losses are scarce, but they indicate that patients with severe health anxiety report more disability days compared with a medical outpatient group 42 and the general population 15, 62, 65 and have an increased risk of disability pension 66. A recent systematic review on the economics of medically unexplained symptoms, including health anxiety, stressed that more extensive research on indirect costs as well as long-term perspectives are needed 67.
Treatment of health anxiety
Until the late 1980s, treatment of health anxiety was largely considered unproductive, and in some cases extreme interven- tions such as prefrontal lobotomies were preformed in attempt to reduce the disorder 68.
During the 1990s, theory-driven studies of the effectiveness of cognitive and behavioural therapies lead to major advances in the understanding of both nature and treatment of health anxiety 69. The emphasis on cognitive analyses and concepts helped thera- pists and researchers place the concept of health anxiety on more solid foundation and challenged the earlier predominant view of health anxiety as a treatment-resistant condition.
Patients with health anxiety may show a preference for psycho- logical treatment over pharmacotherapy 70, yet only two studies
47, 71 have examined the effect of pharmacotherapy in health anxiety in RCT. In both studies, pharmacotherapy (respectively Fluoxetine and Paroxetine) was found to be superior to placebo, still in one of the studies 47 including a psychotherapy compari- son, no significant treatment effect was found between pharma- cotherapy and psychotherapy. The long-term effects of pharma- cotherapy on health anxiety remain to be investigated.
Pscychoeducational approaches have also shown effectiveness in the treatment of health anxiety 72-75, though this approach has been suggested adequate predominantly for mild and uncompli- cated health anxiety 76. So far, the most studied and most effec- tive treatment approach is different versions of cognitive behav- ioural therapy (CBT), and in the latest Cochrane review 69 and narrative review 77 of psychological treatments for health anxiety, CBT is recommended as the gold standard intervention for health anxiety. Still, the Cochrane review failed to find any superiority of CBT over non-specific therapies 69, which has been pointed out to be in contrast to reports in many anxiety disorders 78. Recently, a third-wave development of CBT combining traditional cognitive- behavioural strategies with mindfulness (Mindfulness-Based Cognitive Therapy) has been tested on severe health anxiety and likewise shown to be effective in improving severe health anxiety
49. Only, this recent Mindfulness-Based Cognitive Therapy study 49 has tested group therapy in an RCT, whereas earlier trials on health anxiety have focused on individual CBT as group-based approaches have been dismissed as counteractive 79.
Currently 12 RCTs on the treatment of health anxiety 30, 33, 47, 49-51,
80-87 (see Table 2 below for overview of RCTs) have been pub-
lished and generally yield moderate to large effect sizes. Howev- er, earlier trials show methodological shortcomings 69, 77, which may question whether outcomes and discrepancies between studies reflect real effects or are biased by selection of sample (e.g. due to various definitions and diagnostic criteria for classifi- cation) or methods of analysis. Also, recent studies have drawn attention to the importance of transparent and easily identifiable diagnostic criteria, thus warranting a more homogeneous patient group 48, 50.
Reviews 69, 77 have pointed out methodological issues with earlier trials such as; a large proportion of potentially eligible patients declining participation, no use of power calculations, limited use of standardised diagnostic instruments, limited independent assessor ratings, high drop out rates, and no or short follow-up periods.
Some patients with health anxiety do not respond to CBT inter- ventions 34, and many who show improvement do not maintain their gains long-term 30. Two earlier follow-up studies on medical outpatients both found that 2/3 of patients still met at least some diagnostic criteria for hypochondriasis after a follow-up period of 1 and 4 to 5 years 41, 42. Newly published results from a 6-year follow up CBT study confirmed earlier findings of 2/3 of patients maintaining case status 30. In this way, alternatives to traditional (individual) CBT approaches need exploring in order to improve treatment results 69
A new treatment approach for health anxiety - Acceptance and Commitment Therapy
Acceptance and Commitment Therapy (ACT) 88, 89 is part of a new generation of behavioural therapies that combine well-known behaviour techniques from CBT with strategies that promote acceptance, that is intentionally allowing painful psychological events such as illness worry, to be present and felt in order to be able to move in a valued direction 90. In therapy, ACT combines acceptance- and mindfulness-based processes with behavioural strategies to increase the psychological flexibility of the individu- al. Psychological flexibility spans a wide range of human abilities and can be defined as the ability to recognize and adapt to vari- ous situational demands in the present moment and without needless defence, in the service of chosen values, even when difficult thoughts, feeling or sensations are present 91. The oppo- site position - psychological inflexibility – is in ACT thought to emerge from experiential avoidance, which refers to rigid and in the long term fruitless attempts to avoid or gain control over private events such as aversive thoughts or bodily sensations.
Behavioural patterns dominated by experiential avoidance may be problematic as they restrict behaviour and hence may result in lower quality of life due to dominance of rule-driven versus val- ues-driven behaviour.
ACT has an emphasis on the function of inner experiences, that is how thoughts are experienced and regulated, rather than on testing the validity, form, intensity or frequency of such experi- ences. In this way, the ultimate goal of ACT is to increase psycho- logical flexibility and hence strengthen the ability to act in accord- ance with personal values even in the presence of anxiety. To some extent, ACT uses techniques from CBT, but the goals may differ. For instance, when exposure is applied in ACT-G for health anxiety it is in order to increase behavioural flexibility in the pres- ence of illness worry, that is expand the behavioural repertoire
Table 2. RCTs of treatment for health anxiety
Note 1: * included in Thomson & Pages Cochrane review 69 Note 2: $ included in Boumans narrative review 77.
Note 3: BIB: Bibliotherapy/booklet; BSM: behavioural stress management; BT: Behavioural therapy; CAU: care as usual; CBT: cognitive behavioural therapy; CT: cognitive therapy; DSM: Diagnostic and Statistical Manual of Mental Disorders; ET: explanatory therapy;
(S)HAI: (short) Health anxiety inventory; IAS: Illness Attitudes Scale; MBCT: Mindfulness-based cognitive therapy; STPP: short-term psychodynamic psychotherapy; TAU: treatment as usual; US: usual services (uunrestricted services); VAS: visual analogue scales (de- vised by the authors to assess different aspects of health anxiety); WL: waitlist;.
previously narrowed in the presence of worry, whereas in tradi- tional CBT the goal may primarily be to reduce the arousal evoked by illness worry. In this way, CBT’s focus on modification of the content of cognition, such as dysfunctional thoughts, in order to influence behaviour and emotion may differ from ACT’s focus on increasing awareness of and changing the function of inner expe- riences 90.
Though there are fundamental differences between CBT and ACT as regards both view of psychopathology and focus in treatment, it has been argued that therapeutic techniques used in ACT are compatible with CBT, and that the added acceptance-based tech- niques may improve outcome in many disorders 92.
The empirical support for ACT has increased substantially in re- cent years with positive results for an array of problems 93-101. Even so, a recent systematic review emphasises that methodolog- ically rigorous trials of ACT are very much needed 102, 103. A litera- ture search for ACT and health anxiety found no RCTs when initi- ating this study in 2009, and the status is today unchanged except for the trial of this thesis (the intervention is described in paper II). However, to date there are almost 100 RCTs on ACT with promising evidence for effectiveness in the treatment of both anxiety and depression 104.
AIMS OF THE STUDY The aims of the thesis were:
To develop, test the feasibility and effect of a new treatment approach, Acceptance and Commitment Therapy in groups (ACT- G), for patients with severe health anxiety:
I. To test the feasibility and acceptance of ACT-G (Paper I) II. To evaluate the effect of ACT-G in a randomised con- trolled trial as well as the acceptance of the diagnostic re- search criteria for health anxiety (Paper II)
To explore the association between severe health anxiety and sick leave:
III. To assess
a) the association between severe health anxiety and sickness-related benefits compared with a matched general population sample during a 5-year period before randomisation to ACT-G, and
b) the treatment effect of ACT-G on sickness-related benefits in patients with severe health anxiety (Paper III).
GENERAL DISCUSSION OF METHODS
I will discuss methodological issues with regard to the design and data sources of the study. In the discussion section of the individ- ual papers, general strengths and limitations of the study have been stated, and therefore, to avoid recapitulation of arguments provided, not all of these will be repeated in the present discus- sion of methods.
Reviews of earlier RCTs on health anxiety 69, 77 have focused on a number of methodological shortcomings in existing trials. Based on these, a number of methodological topics have been chosen for discussion with the purpose of shedding further light on the contribution of the present results to the current body of knowledge as well as the generalisability of the results of the present study.
Design
Recruitment and treatment setting
The generalisability of findings from the present study to individ- uals in the general population with severe health anxiety can be affected by multiple factors. In the present study, we set up ex- clusion criteria in order to ensure that included patients were able to participate in the group therapy and that their symptoms were not better explained by another disorder, hence we only included patients of Scandinavian origin between 20 and 60 years of age, who did not have a drug or alcohol abuse, were not preg- nant and did not have a history of severe psychiatric morbidity such as psychotic and bipolar disorders. For ethical reasons, i.e. to ensure we only included patients severely disabled by health anxiety and with strong treatment need, we only included pa- tients reaching diagnostic criteria for severe health anxiety, which implies moderate to severe impairment 1.
In Denmark, almost all health care is free of charge for all citizens, including visits to GPs, and is therefore easily accessible for every citizen. Also, we had patients referred from both rural and urban parts of western Denmark (catchment area of approximately 2.5 million persons).
Up to and during the recruitment phase of the study, we sent thorough information material to all GPs in West Denmark in order to inform GPs on the health anxiety diagnosis and criteria for referral. In this way, we tried to reduce potential selection bias as all GPs received the same information and hence all pa- tients with severe health anxiety had a theoretical chance for referral to the study. In this way, the representativeness of the study sample may be fair to good compared with the general population of Danish adults with severe health anxiety, with the exception of the gender representation. In the present study, as in all other RCTs on health anxiety, predominantly women were referred and included in the study. As earlier studies have found no gender differences in the prevalence of health anxiety 1, 15, 44-
46, the inclusion of predominantly women in this and other stud- Background at a glance
The issues of classification have hampered research in health anxiety and hence have added a spurious com- plexity to our understanding of treatment.
There is a need to establish meaningful boundaries of caseness for health anxiety.
Severe health anxiety tends to show the highest preva- lence in medical settings.
Health anxiety may be moderately heritable yet more strongly influenced by environmental factors.
Health anxiety may have an early onset and a chronic course in severe cases left untreated.
Overall, persons with severe health anxiety are high users of health care.
Ability to work is an additional outcome of clinical and societal significance, but the association between health anxiety and sick leave is scarcely investigated.
Research has shown effectiveness of CBT, yet many studies have been affected by methodological problems
Further development and improvement of treatment approaches may help enhance treatment results.
ies may potentially affect the outcome and to some extent un- dermine the external validity. It may be that women respond differently to the treatment compared to men. We can not say why predominantly women were included in our trial. The reason may be that women in general consult their GP more often than men 105 and thus are overrepresented in general practice. Also, presentation of health anxiety could be gender-specific and per- haps more easily identifiable in women thus making GPs more aware of health anxiety in women. In order to correct for this bias and equally distribute the under-represented men, we stratified for gender in the block-randomisation.
Also, we can not know if ACT-G for instance has attracted espe- cially well-educated patients as the large majority of patients in this study, as well as other RCTs on health anxiety, are well- educated. This may well be a characteristic for patients with severe health anxiety too. Moreover, it may be that patients in rural areas have been less likely to accept a referral due to trans- portation time to the clinic.
At large, differences in for instance recruitment procedure and setting complicates comparison of results across studies. In the present study, patients were recruited from mainly GPs, yet a very small proportion of patients were referred from secondary care. In the latter cases, the GP was informed of the referral.
Accordingly, the study sample consisted of patients all consulted by a medical doctor, which, compared to other studies recruiting through self-referrals 47, 49-51, 82, may have ensured a certain amount of impairment present and segregation of primary physi- cal diseases, and hence a more homogenous sample compared to self-referred samples. One may expect this to result in a more severely ill sample in the present study compared to other stud- ies, although comparing this study’s baseline scores on illness worry and illness duration with other RCTs does not seem to suggest this.
Furthermore, the present study was conducted at a specialised clinic for functional disorders, within a psychiatric specialty, yet located at a general hospital. In other RCTs on health anxiety treatment, the study was conducted at secondary care settings 82,
84-87 and/or university clinics 49, 80. It may be that some patients have declined a referral from their GP because the study and treatment took place at a specialised clinic and in this way only patients who accepted the referral to this type of setting were included. Despite this, patients have generally accepted referral, and according to the feedback from patients and GPs, they have welcomed the treatment.
Decline of participation and dropout
In the present trial we experienced that very few (9%, 24 of 254) referred patients and only 5% (9 of 173) of the potentially eligible patients declined participation in the study. Furthermore, only 6%
(4 of 63) dropped out of treatment. Compared to some other trials having been hampered by large decline (70-80%) 82, 86 and dropout of up to 1/3 of patients 47, the present RCT shows low numbers of decline and dropout.
There can be different explanations for these findings, potential ones might be; 1) GPs have only referred a selected type of pa- tients, 2) lack of available treatment alternatives 3) the specific treatment offered, or 4) the assessment procedure at the re- search clinic. Firstly, it is possible that GPs only referred specific cases, such as the most severe and troublesome, the most moti- vated, or only those open-minded towards psychotherapy. In this way, a predominantly motivated sample may have been referred.
It seems unlikely that GPs would have intentionally made this selection as the information material specifically asked for refer-
rals of all patients with severe illness worry according to in- and exclusion criteria, still some patients may have declined a referral.
Also, we did experience some decline of participation among referred patients (9%, 24 of 254). A second possible explanation is that the low decline and dropout reflects the lack of alternative treatment offers. This may be the case, yet it may not explain the difference in decline and dropout seen in the present study com- pared to others as a lack of treatment alternative has been the same for other studies and still is the case in most countries. A third possible explanation may be that the specific treatment offered – a group therapy with an acceptance-based focus – has been very attractive and meaningful to the patients. This explana- tion might be supported by the fact that besides low dropout, a high adherence to treatment was found as patients in ACT-G had a median of attendance of nine sessions (interquartile range (IQR) 8-10). The only other group-based RCT 49 for health anxiety also found a very low dropout rate. Furthermore, this treatment, the same way as the present, was based on a further development (third-wave) of CBT.
Another explanation for the low decline and dropout may be that assessors at the research clinic were very motivating in getting patients to participate. As only 5% (9 of 173) declined participa- tion after assessment, this could well be the case. Yet, this may not be the only explanation as one would expect only an immedi- ate high-motivation effect and a later high dropout of treatment, but this was not seen. Based on our clinical experience, we think that the low decline and dropout may at large be caused by a high patient acceptance of the diagnosis, due to the easily identifiable and empirically validated diagnostic criteria, and a thorough assessment. At assessment, patients received psychoeducation on e.g. the nature and cause of health anxiety and received a diagnosis, which the large majority (97-98%) categoried as the right diagnosis to fit their ailment and agreed that the diagnosis helped them to better understand their symptoms 4. Other stud- ies 47, 49-51, 80, 86, 87 have at large used the broadly defined DSM-IV hypochondriasis diagnostic criteria, and this may first of all have caused decline in participation as patients often find the diagnosis stigmatising and maybe can not identify with it. Secondly, it may have recruited a more heterogeneous sample, of which some might not have found the treatment focus on illness worry suita- ble for their condition and thus have dropped out.
Diagnostic criteria for inclusion
In this study, we used well-defined empirically established diag- nostic criteria for health anxiety 1. Clinically, we experienced that these criteria were transparent, identifiable and easily under- stood by the patients and the referring GP compared with the more poorly defined DSM-IV or ICD-10 criteria 1. Also, the pa- tients easily accepted the new diagnosis, which they contrarily to the DSM-IV and ICD-10 diagnostic labels did not find stigmatising.
Furthermore, the DSM-IV and ICD-10 diagnoses have been criti- cised for being too broad and hence show difficulties in differen- tiating health anxiety from other somatoform disorders and re- sulting in heterogeneous samples 1, 15. Furthermore, in the diagnostic criteria for health anxiety, the DSM-IV B criterion re- garding “appropriate medical evaluation and reassurance” has been omitted, which may generate a more homogeneous sample.
According to the diagnostic criteria used for the present trial, patients with primary severe other somatoform disorders were excluded such as a patient for whom the primary concern and diagnosis was fibromyalgia and where severe illness worry pre- sented secondary. The diagnostic criteria may thus have helped
specify a sample of patients that only represents a subgroup of the participants included in other trials. It may be that differences found between the current study and other studies as regards treatment effect might to some extent be explained by different samples due to use of different definitions and assessment meth- ods.
Waitlist as a control condition
A waitlist control group can be defined as a group of patients who have been assigned to wait for treatment and are aware that they are currently not receiving the treatment they are waiting for.
First of all, in that way, a waitlist control group is not untreated. In the present trial, patients were referred, assessed, measured, diagnosed, received psychoeducation, and were randomised.
After the assessment, the GP and the patient received a prelimi- nary discharge letter regarding diagnosis and illness history. In the present RCT, both groups improved statistically significantly in illness worry from pre-screening to randomisation, which may be due to the assessment. Patients on the waitlist sustained this improvement during the 10-month follow-up period, but they did not improve further on the primary outcome of illness worry (SRM randomisation to 10-month follow-up = -0.07, 95% CI 0.20- 0.33) 4.
Secondly, it has been suggested 106 that patients in waitlist control groups improve less or not at all as they know they are awaiting treatment and efforts at improving are interrupted. Also, the control group knows that they are not yet receiving active treat- ment and has no reason to expect positive change, also referred to as treatment expectancy effects. Thus, the waitlist design may have inflated the reported effects of ACT-G. Being assigned to the waitlist group may have reduced participation in other beneficial activities during the 10-month period, even though such partici- pation was not discouraged.
On the other hand, it is possible that at least some patients allo- cated to waitlist, after having received a brief introduction to ACT and mindfulness as part of the assessment, may have acquired self-help books on ACT or have felt motivated to contact a thera- pist in the wait period. In this way, the control group may have been contaminated, and the effect of the intervention could be underestimated. In the present trial, patients on waitlist received usual care by their GP, and there were no restrictions applied to the psychological or pharmacological interventions or on referrals to secondary care or mental health services during this time period. We can not know what impact the waitlist design may have had on the effect size, yet we chose not to include the signif- icant effect of the assessment in the calculated effect sizes, and therefore one may expect the effect of the intervention to be even larger in everyday clinical practice.
In a recent meta-analysis 107 of CBT treatment outcomes for health anxiety is was found that waitlist control conditions (4 studies) showed larger effect sizes than treatment as usual condi- tions (5 studies). In the review it was not possible to examine other control conditions (e.g. psychological or pill placebo control conditions) due to few available studies (3 studies). Considering the possibility that a waitlist design may overestimate treatment effects, as suggested in the recent meta-analysis on treatment of health anxiety 107 and in RCTs in general 106, this must be taken into consideration when comparing effect sizes with other trials, and the findings of the present study need replication in studies with an active control group.
The present study was a somewhat pragmatic trial, which is why we - for ethical reasons - decided on the waitlist design offering
the control group treatment after the follow-up period, which the large majority of the patients accepted (84% 53 of 63). We con- sidered offering the control group weekly check-in sessions, which would serve as someone showing concern, but as many patients would have to spend up to two hours on transportation each way to the clinic, we found it unethical and unfeasible. We think that the waitlist design was the best choice as no other specialised treatment was available for this patient group in Denmark at that time, just like in most other countries. With no other specialised treatment available for health anxiety, it is difficult to set up a treatment as usual condition.
Data
Processing of questionnaire data
Questionnaires were designed and processed using the TELEform software program, which allows for optical reading and hence has showed low error rate 108. A research secretary, student and data manager scanned the questionnaires following thoroughly prede- fined guidelines on how to handle and document cases of doubt.
Project head and statistician were responsible for the further collation of data.
Self-rated measures of health anxiety
In the area of health anxiety, predominantly three questionnaires have been used as self-rated measures of health anxiety; the Illness Attitudes Scales (IAS), the Health Anxiety Inventory (HAI) and the Whiteley Index (WI). The validity and reliability of measures are often based primarily on the discriminant validity such as the sensitivity to differentiate between for instance cases of health anxiety and other somatoform disorders and the sensi- tivity to changes of treatment. All questionnaires may be useful as screening instruments for preliminary case identification, but can not be used alone for the purpose of diagnosis 109.
The HAI short version consists of 16 items with each a group of 4 statements. In a validation study, the scale was found to have good discriminant validity, good test-retest reliabilities and also found to be sensitive to treatment effects 110.
The IAS measures attitudes, belief and fears associated with health anxiety and consists of 27 items distributed within 9 scales, of which only the first 7 directly concern fears and beliefs of health anxiety 111, 112. The 9 scales are clinically-derived and show good test-retest reliabilities.
Both the IAS and HAI show good discriminant validity and sensitiv- ity to changes of treatment, yet may be considered quite com- prehensive instruments for screening purposes.
The WI 113 was developed in order to clarify the symptom clusters that are seen in clinical health anxiety by using factor analysis.
The WI was developed by Pilowsky et al. almost 50 years ago and originally consisted of 14 items with dichotomous answer catego- ries (true-false). Hiller et al. 114 investigated the similarity between IAS and WI and found that the two instruments were highly corre- lated (0.80) both yielding high sensitivity/specificity (71-80%), yet the IAS showed superior discriminative validity. The authors stressed that the nine original scales of the IAS are not sufficiently empirically supported. Also, a recent qualitative review of the dimensional assessment of health anxiety 115 concluded that the WI is one of the most validated and used instruments, but that it should be a second choice to the IAS as the IAS shows superiority over the WI in clinimetric properties, in particular sensitivity to treatment-related changes and content validity. However, it should be stressed that both the qualitative review and the study by Hiller et al. were based on assessing the 19-item version of the WI with a dichotomous response format.
The WI has since the introduction of the original version been refined and psychometrically rigorously tested compared to other instruments e.g. by means of a latent trait model (the Rasch model) 116 to a 7-item version with a Likert-scoring from 1 to 5.
Using a Likert scale provides a wider response range and may hence allow for a better representation of the continuum of health anxiety 117. The WI-7 has been widely used and has demonstrated satisfactory psychometric properties in primary care samples, with high internal validity and impressive external validity for screening DSM-IV somatisation disorder and hypo- chondriasis/health anxiety 116, 118. Furthermore, the WI has been shown to have a satisfying responsiveness to changes over time
119.
In the present study, the WI was chosen because of its sound psychometric properties and simplicity, which makes it attractive as a screening instrument, and the primary outcome was decided a priori as the mean change in the WI-7 score 116 from baseline (randomisation) to the 10-month follow-up (6 months after treatment).
In all questionnaire data, there is the risk of recall bias as a possi- ble confounder, though on the WI patients were asked to assess how much they had been bothered by illness worry within the last four weeks, which is a fairly short time period hopefully min- imising recall bias. Still, assessing illness worry over a period of 4 weeks may serve as a potential bias as illness worry clinically is known to fluctuate over short time spans for some patients de- pending on how much or how little they are confronted with illness worry evoking information or sensations. Fluctuations may cause problems both in respect to validity and to estimation of effects.
It is up to future research to further refine the WI. It may be that the discriminant validity of WI could be heightened by taking out the two items in the 7-item version concerning complaints of general pain/aches and multiple symptoms, which may not un- derlie health anxiety specifically, but somatisation or somatoform disorders in general. A recent study 120 has used data from a large population-based study to conduct confirmatory factor analysis and item response theory analysis of WI and found evidence for a 6-item (item concerning different pains and aches taken out), single factor model of WI, yet these findings need to be replicated in other samples.
Cut-off for identification of case status
In this study we used the WI both as a dimensional measure (scale score 0-100) and as a categorical measure (dichotomization in regards to cut-off of <21.4) to establish potential case status at 10 months follow-up. The cut-off determining clinical case status was based on existing data on patients with severe health anxiety
2, 5. Still it may be said to be a somewhat arbitrary cut off and the validity of case status based on just a cut off is always questiona- ble. Conradt et al.’s 118 has suggested that for screening purposes, a cut-off score of ½ on the WI shows enough sensitivity, whereas a cut-off score of 2/3 (equalling a WI <33.33 on a 0-100 score scale) shows the best balance for sensitivity and specificity for identification of cases of health anxiety. If we had applied the 2/3 cut off to estimate case status at end point in the present study, 58% (30 of 52) of patients in ACT-G and 24% (13 of 55) in the waitlist were no longer clinical cases of health anxiety. Compara- bly, the more conservative cut-off chosen in our study proposed a considerably lower proportion of 27% (14 of 52) of non-cases in ACT-G and 9% (5 of 55) in the waitlist.
Using less restrictive criteria, such as a questionnaire, to assess caseness, poses a risk of under- or overestimating prevalence of a disorder and missing important symptom patterns. Furthermore, a questionnaire is often not capable of differentiating whether symptoms are better explained by another disorder, e.g. major depression or an anxiety disorder. Health anxiety may be seen on a continuum of severity ranging from transient and relatively mild symptoms that may be difficult to separate from normal physio- logical phenomena, and to conditions of severe, chronic and disabling health anxiety. In all cases, a diagnostic reassessment with a (semi-) structured interview would be preferable to a somewhat arbitrary cut-off on a questionnaire. A questionnaire can be used as a screening instrument, but can not be used to diagnose, and in studies where this is done, it might show a ten- dency to include many cases without clinical relevance 109. Creed
& Barsky 19 refer in a review of the epidemiology of somatisation disorders and hypochondriasis to researchers having lowered the threshold of a questionnaire cut-off and hereby increased the number of people reaching case status, but still the additional individuals were no less disabled than those fulfilling the criteria of full ICD-10 hypochondriasis. The authors conclude that future work is still needed to determine the optimal cut-off point in order to determine cases of illness worry that is associated with impairment.
The diagnostic assessment
To classify health anxiety, we used a modified version of the semi- structured psychiatric interview, Schedules for Clinical Assess- ment in Neuropsychiatry (SCAN) which includes the symptoms of health anxiety used in the research diagnostic criteria of health anxiety 1, 121, 122. SCAN is symptom-driven, contrary to diagnosis- driven, and uses a ‘bottom-up’ approach, which is why it is not limited to specific symptoms included in diagnostic criteria for various conditions, and therefore its’ validity does not rely solely on a diagnostic system. Nearly any DSM-IV or ICD-10 diagnoses can be established based on the SCAN diagnostic algorithms. In our study, the SCAN assessments were carried out in a clinical context by six certified SCAN assessors who were all trained clini- cians, implying that assessment at large is comparable to other diagnostic situations such as clinical consultations, which is con- sidered a strength of the SCAN. In the study, we used SCAN for case identification, which we find is a strength of the study and may have increased the diagnostic consistency and validity of health anxiety, whereas others 82, 85 assessed cases of health anxiety based on questionnaire data alone or layman interviews.
Compared to questionnaires alone, this semi-structured interview allows for qualitative aspects such as impairment and well-being and assesses possible organic causes for symptoms. In this way, using SCAN may have allowed us to overcome some of the meth- odological problems in determining the origin of symptoms and whether severity is clinically significant.
In order to minimise potential problems of misclassification, the interviewers were: 1) instructed to consult relevant medical spe- cialists in case of doubt of symptom origin, 2) instructed to go through all chapters in SCAN, including detailed symptom descrip- tion, in case of the slightest doubt of the origin of the reported symptoms in order to rule out another primary diagnosis that may better account for the reported symptoms. Also, throughout the inclusion period, video-taped SCAN interviews were randomly selected to assess inter-rater agreement on diagnoses.
